Narrow Results

To study the data association, a structure called a hierarchy tree is constructed. It is based on the approach to hierarchical data processing, and constituted by different level partitions of a data set. This leads to the definition of the data association, thereby links two hierarchy trees together. The research on the data association focuses on the way to check whether data are associated with other data. The investigation includes the issues: the intuitive and formal methods for constructing hierarchy trees, the technique of making granules hierarchical, the sufficient and necessary condition for measuring the data association, the analysis of basing the closer data association on the closer data identity, the discussion of connecting numerical information with association closeness, etc. Crucially, the hierarchical data processing and numerical information are important characteristics of the research. As an applied example, two hierarchy trees are set up, demonstrating the hierarchical granulation process of two actual data sets. Data associations between the data sets are characterized by the approach developed in this paper, which provides the basis of algorithm design for the actual problem. In particular, since the research is relevant to granules and alterations of granularity, it may offer an avenue of research on granular computing.

Visual surveillance in wide areas (e.g. airports) relies on sparsely distributed cameras, that is, cameras that observe nonoverlapping scenes. In this setup, multiobject tracking requires reidentification of an object when it leaves one field of view, and later appears at some other. Although similar association problems are common for multiobject tracking scenarios, in the distributed case one has to cope with asynchronous observations and cannot assume smooth motion of the objects. In this paper, we propose a method for human indoor tracking. The method is based on a Dynamic Bayes Network (DBN) as a probabilistic model for the observations. The edges of the network define the correspondences between observations of the same object. Accordingly, we derive an approximate EM-like method for selecting the most likely structure of DBN and learning model parameters. The presented algorithm is tested on a collection of real-world observations gathered by a system of cameras in an office building.

Herein, a novel methodology is proposed for real-time recognition of human activity in a compressed domain of videos based on motion vectors and self-attention mechanism using vision transformers, and it is termed as motion vectors and vision transformers (MVViT). The videos in MPEG-4 and H.264 compression formats are considered for this study. Any video source without any prior setup could be considered by adopting the proposed method to the corresponding video codecs and camera settings. Existing algorithms for recognition of human action in a compressed video have some limitations in this regard, such as (i) requirement of keyframes at a fixed interval, (ii) usage of P frames only, and (iii) normally support single codec only. These limitations are overcome in the proposed method by using arbitrary keyframe intervals, using both P and B frames, and supporting MPEG-4 as well as H.264 codecs. The experimentation is carried out using the benchmark datasets, namely, UCF101, HMDB51, and THUMOS14, and the recognition accuracy in a compressed domain is found to be comparable to that observed in raw video data but at reduced cost of computation. The proposed MVViT method has outperformed other recent methods in terms of a lesser (61.0%) number of parameters and (63.7%) Giga Floating Point Operations Per Second (GFLOPS), while significantly improving accuracy by 0.8%, 5.9% and 16.6% for UCF101, HMDB51 and THUMOS14, respectively. Also, it is observed that the speed is increased by 8% in case of UCF101 when compared to the highest speed reported in the literature on the same dataset. The ablation study of the proposed method has been done using MVViT variants for different codecs and the performance analysis is done in comparison with the state-of-the-art network models.

In order for a service robot to approach humans and provide the services it has been designed for, an efficient system for people tracking and identification must be developed. This paper presents a novel solution to the problem that makes use of different sensors and data fusion techniques. The robot utilizes a laser device and a PTZ color camera to detect, respectively, human legs and faces. The relative information is integrated, in real-time, using a sequential implementation of Unscented Kalman Filter. Furthermore, thanks to an histogram comparison with a measure based on the Bhattacharyya coefficient, people are also identified and labelled according to their clothes. This measure is also used to improve the robustness of the data association process. The effectiveness of the proposed method is shown by experiments with a real mobile robot in challenging situations.

A massive amount of video data is recorded daily for forensic post analysis and computer vision applications. The analyses of this data often require multiple object tracking (MOT). Advancements in image analysis algorithms and global optimization techniques have improved the accuracy of MOT, often at the cost of slow processing speed which limits its applications only to small video datasets. With the focus on speed, a fast-iterative data association technique (FIDA) for MOT that uses a tracking-by-detection paradigm and finds a locally optimal solution with a low computational overhead is introduced. The performance analyses conducted on a set of benchmark video datasets show that the proposed technique is significantly faster (50–600 times) than the existing state-of-the-art techniques that produce a comparable tracking accuracy.

To make up for the defects of semanteme expression about linked data, this paper proposes a semanteme expressing method of associated entities based on relationship diagram so as to realize the machine expression and recognition of associated semanteme in relational databases. Starting with the structure and relationship of relational schema, this paper analyzes the rich semanteme of associated entities and presents the semanteme parsing method based on the traversal path as well as its formal expression; the analysis of instance database is also carried out. Studies show that this method can comprehensively parse and express the associated semanteme of entities. This work has reference significance for the research of intelligent semanteme synthesis and for semanteme-oriented intelligent query.

This paper studies the problem of data association in multiple maneuvering targets locating and tracking. Genetic algorithm is used successfully in solving the complex optimization and the industrial engineering problem. Recently researching on genetic algorithm has attracted a lot of attention. This paper puts forward a way of using Genetic Algorithms to resolve the problem of multiple target data association. The simulation results show that the algorithm used in this paper is able to avoid the local extremum and the outcome is satisfactory.